CVD diamond single crystals with NV centres: a review of material synthesis and technology for quantum sensing applications

HAL Id: hal-03052857

https://hal.archives-ouvertes.fr/hal-03052857

Submitted on 10 Dec 2020

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

CVD diamond single crystals with NV centres: a review of material synthesis and technology for quantum

sensing applications

J Achard, V Jacques, Alexandre Tallaire

To cite this version:

J Achard, V Jacques, Alexandre Tallaire. CVD diamond single crystals with NV centres: a review of

material synthesis and technology for quantum sensing applications. Journal of Physics D: Applied

Physics, IOP Publishing, 2020, 53 (31), pp.313001. �10.1088/1361-6463/ab81d1�. �hal-03052857�

CVD diamond single crystals with NV centres: a review of material synthesis and technology for quantum sensing applications

J. Achard

, V. Jacques

and A. Tallaire

1

LSPM, Université Paris 13, Sorbonne Paris Cité, CNRS, 99, avenue JB Clément, 93430 Villetaneuse, France

2

L2C, Laboratoire Charles Coulomb, Université de Montpellier and CNRS, 34095 Montpellier, France

3

IRCP, Ecole Nationale Supérieure de Chimie de Paris, 11, rue Pierre et Marie Curie, 75005 Paris, France

* Corresponding author: jocelyn.achard@lspm.cnrs.fr

Abstract:

Diamond is a host for a wide variety of colour centres that have demonstrated outstanding optical and spin properties. Among them, the nitrogen-vacancy (NV) centre is by far the most investigated owing to its superior characteristics, which promise the development of highly sophisticated quantum devices, in particular for sensing applications. Nevertheless, harnessing the potential of these centres mainly relies on the availability of high quality and purity diamond single crystals that need to be specially designed and engineered for this purpose. The plasma assisted Chemical Vapour Deposition (CVD) has become a key enabling technology in this field of research. Nitrogen can indeed be directly in-situ doped into a high crystalline quality diamond matrix in a controlled way allowing the production of single isolated centres or ensembles that can potentially be integrated into a device.

In this paper we will provide an overview on the requirements for synthesizing “quantum-grade” diamond films by CVD. These include the reduction of impurities and surrounding spins that limit coherence times, the control of NV density in a wide range of concentrations as well as their spatial localization within the diamond. Enhancing the charge state and preferential orientation of the colour centres is also discussed.

These improvements in material fabrication have contributed to position diamond as one of the most

promising solid-state quantum system and the first industrial applications in sensing are just starting to

emerge.

1. Introduction

Solid-state quantum systems that possess long-lived spin and/or optical coherence are likely to play key roles in the development of a broad range of applications in quantum technologies (QT), from quantum networks, to information processing and quantum sensing [1]. In this context, various candidates are being considered such as superconducting circuits [2], donors in silicon [3, 4], rare-earth ions in oxide crystals [5], quantum dots [6] or defects in semiconductors [7]. Among them, colour centres in diamond are arguably one of the most promising and studied systems [8]. The growing interest that surrounds this material essentially stems from the outstanding optical and spin properties of the nitrogen-vacancy (NV) colour centre [9], which have opened up a plethora of potential breakthrough applications in QTs, including the first demonstration of kilometer-scale entanglement between solid-state spin qubits for quantum networks [10], the realization of quantum error correction protocols [11], and the development of highly sensitive quantum sensors [12- 14], which are already close to commercial products [15]. In addition, the progresses achieved in harnessing NV centres’ potential for QTs has fostered the emergence of other colour centres in diamond, particularly those of group IV, that exhibit complementary properties such as SiV, GeV, SnV, or PbV [16-20]. At the heart of the success of diamond as a platform for QTs, are the fundamental science and the technologies that have allowed the fabrication of specially designed and engineered “quantum grade” synthetic crystals.

Indeed, most practical demonstrations and advances in diamond-based QT leverage on material development.

Both the High-Pressure High-Temperature (HPHT) and Chemical Vapour Deposition (CVD) techniques are currently used to fabricate diamonds with optimized properties for industrial and high-tech applications.

These synthetic diamond growth technologies have witnessed tremendous improvements over the past

decades leading to ever thicker, larger and higher purity crystals [21]. Gem-quality material with fancy colours

or colourless and up to several carats in size have been obtained, which may be seen by some, as a threat to

the stability of the natural diamond market established for jewellery [22]. One of the driving forces for

innovation however, has been the field of electronics in which diamond detectors as well as power devices

are regarded as technologically disrupting with outstanding figures of merit [23]. Schottky diodes and field

effect transistors are foreseen to allow the operation of smaller components that can drive exceptionally

high currents and sustain high voltages in harsh environments [24-26]. While these devices are still in their

infancy, their development has required improvements in the synthesis of fairly thick diamond films (several

hundreds of µm) with a purity down to the ppb (parts per billion) level and a surface area as large as possible

to facilitate processing and integration [27]. To this end, electrical doping using boron for p-type [28] and

phosphorous for n-type [29] has been explored. Nitrogen which is a deep passivating donor with an activation

energy of about 1.7 eV [30] generally needs to be avoided. However, like most wide band gap

semiconductors, diamond suffers from a high activation energy of its dopants and asymmetric doping with

n-type being extremely difficult to achieve on a standard (100) orientation and requiring non-conventional

growth conditions and substrates [31-33]. The enormous progresses made in this area during the last

decades have played a crucial role in unleashing the potential of this material for QTs and have contributed

to make diamond material available to this broad research community. In fact, an accurate control over the

amount of residual impurities (such as nitrogen and boron), isotopic carbon content as well as crystalline

defects that have a deleterious effect on spin coherence times are key to material adoption in QTs. While

HPHT can produce bulk crystals with high crystalline perfection, purity remains limited and the technique is

usually not flexible enough to allow for a precise engineering of “quantum grade” layers of material even if,

as it will be discussed in part 3.1, some HPHT diamonds have been studied and exploited for QT

demonstrations. The route that is thus most widely followed is to homoepitaxially grow a thin diamond film

with optimized properties by CVD on a HPHT diamond substrate possessing appropriate crystalline quality and orientation. Although the CVD growth method is relatively mature, crystalline films that can deliver optimized performance in QTs are yet not routinely produced.

In this review we discuss some of the achievements and the remaining challenges that are crucial to the highly demanding field of diamond-based QTs, with a focus on quantum sensing and imaging applications with NV colour centres. In particular we give special emphasis to the material fabrication through the now well-established CVD technique and focus on in-situ doped material with NV centres for sensing devices.

Issues with other colour centres (SiV, GeV etc.) might be raised but will not be discussed in great details.

2. Material requirements for NV-based quantum sensing applications

In this section, we first identify some key challenges in diamond growth to optimize the performance of quantum sensing applications based on NV colour centres, starting with a brief reminder of their main optical and spin properties.

The NV colour centre consists of a substitutional nitrogen atom (N) combined with a vacancy (V) in a neighboring lattice site of the diamond crystal (figure 1(a)). This point-like defect gives rise to localized electronic states with energy levels deeply buried inside the bandgap of diamond. As a result, the NV centre can be considered as an artificial atom, mostly decoupled from the valence and conduction bands of the host material. Like many point defects in semiconductors, the NV colour centre can be found in various charge states having very different optical and spin properties [9]. Applications in QTs mostly rely on the negatively- charged state (NV

), for which an additional electron is provided by a nearby donor impurity, thus leading to a quantum system with two unpaired electrons. The NV

colour centre exhibits a perfectly photostable photoluminescence (PL) emission with a zero-phonon line at 1.945 eV ( 

= 637 nm), and provides a spin triplet ground level, which can be initialized by optical pumping, coherently manipulated with long coherence time through microwave excitation, and readout by pure optical means (figure 1(b)) [9]. As explained below, these properties are at the heart of NV

based quantum sensing. However, the NV defect can also be stabilized in a positively-charged configuration (NV

), which is optically inactive [34, 35], and more often in a neutral form (NV

), which is characterized by a shift of the zero-phonon line to 2.15 eV (

= 575 nm), and does not feature the appealing spin properties of its negatively charged counterpart [36-38]. As a result, a first requirement on diamond crystals for QT applications is to provide an environment promoting the stabilization of the NV

charge state. In the following, we focus on the spin properties of the NV

configuration, which will be simply referred to as NV for clarity purpose.

A key feature of the NV colour centre is that its ground level is a spin triplet state, S = 1, whose

degeneracy is lifted by spin-spin interaction into a singlet state of spin projection m

=0 and a doublet m

=±1,

separated by 2.87 GHz in the absence of magnetic field (figure 1(b)). Here m

denotes the spin projection

along the NV defect quantization axis, corresponding to a [111] crystal axis joining the nitrogen and the

vacancy. Radiative transition selection rules associated with the spin state quantum number lead to an

efficient polarization of the NV defect in the ground state spin level m

=0 by optical pumping. Furthermore,

the NV defect PL intensity is significantly higher when the m

=0 state is populated. Such a spin-dependent PL

response enables the detection of electron spin resonance (ESR) on a single defect by optical means. Indeed,

when a single NV defect, initially prepared in the m

=0 state through optical pumping, is driven to the m

=±1

spin state by applying a resonant microwave field, a drop of the PL signal is observed, as depicted in

figure 1(c).

Figure 1. (a) Left panel: Optical image of a high-purity diamond crystal grown by CVD. Right panel: Atomic structure of the NV defect.

(b) Simplified energy level scheme. The NV defect is polarized into the spin sublevel m

=0 by optical pumping, and exhibits a spin- dependent photoluminescence (PL) intensity. (c) Optically-detected ESR spectra recorded by monitoring the NV defect PL intensity while sweeping the frequency of the microwave (MW) field. When a magnetic field is applied (lower panel), the ESR transitions are shifted owing to Zeeman effect thus providing a quantitative measurement of the magnetic field projection B

along the NV defect quantization axis.

The first demonstration of optically-detected ESR on a single NV defect was reported in 1997, using a natural diamond sample and a confocal optical microscope operating under ambient conditions [39]. About ten years later, it was shown that these properties can be exploited for the design of a new generation of magnetometers [40-43], providing an unprecedented combination of spatial resolution and magnetic field sensitivity, even at room temperature. Here the magnetic field is evaluated within an atomic-sized detection volume by recording the Zeeman shift of the NV defect’s electron spin sublevels (figure 1(c)), which is given by ∆ =

ℎ

IB

I, where

≈ 28 GHz/T and B

is the magnetic field projection along the NV defect quantization axis. The sensing functionalities of NV defects were then extended to a large number of external perturbations including strain [44], electric fields [45], pressure [46] and temperature [47-49], that all have a direct impact on the ESR frequency. For all these physical quantities, the shot-noise limited sensitivity 𝜂

of a single NV spin sensor scales as [13, 50]

𝜂

∝ 1

𝐶

√𝑅𝑇

(1)

where 𝐶

is the contrast of the optically-detected ESR spectrum, 𝑇

denotes the inhomogeneous spin dephasing time of the NV defect which limits the ESR linewidth, and R is the number of detected photons.

For a single NV defect, the ESR contrast is of the order of 𝐶

≈ 20 %, a value fixed by the intrinsic photophysical properties of the NV defect, which can hardly be modified. The sensitivity can thus be improved either by increasing the collection efficiency of the PL signal [50] or by introducing alternative methods to improve the spin readout fidelity, such as photoelectric detection [51], spin-to-charge conversion [52] or infrared absorption readout [53]. From a material science point of view, the only parameter allowing to optimize the sensitivity is here the spin dephasing time 𝑇

of the NV sensor, which is mainly limited by

valence band

Laser PL

e-spin S=1 conduction band

ms

±1

0

5.5eV

(b)

MW

(a)

1.945eV<latexit sha1_base64="ZjurJECuvVkV83/9l19FMtnMVOw=">AAAC1HicjVHLTsJAFD3UF+Kr6tJNIzFhRYrBKDsSNy4xsUAChLRlwIa+Mp2SEMLKuPUL3OovGf9A/8I7Y0lUYnSatmfOvefM3Hud2PcSYZqvOW1ldW19I79Z2Nre2d3T9w+aSZRyl1lu5Ee87dgJ872QWcITPmvHnNmB47OWM76U8daE8cSLwhsxjVkvsEehN/RcWxDV1/VKuVY9M7rGrMsDgzXnfb1olk21jGVQyUAR2WpE+gu6GCCCixQBGEIIwj5sJPR0UIGJmLgeZsRxQp6KM8xRIG1KWYwybGLH9B3RrpOxIe2lZ6LULp3i08tJaeCENBHlccLyNEPFU+Us2d+8Z8pT3m1KfyfzCogVuCX2L90i8786WYvAEBeqBo9qihUjq3Mzl1R1Rd7c+FKVIIeYOIkHFOeEXaVc9NlQmkTVLntrq/ibypSs3LtZbop3eUsacOXnOJeB<latexit sha1_base64="ZjurJECuvVkV83/9l19FMtnMVOw=">AAAC1HicjVHLTsJAFD3UF+Kr6tJNIzFhRYrBKDsSNy4xsUAChLRlwIa+Mp2SEMLKuPUL3OovGf9A/8I7Y0lUYnSatmfOvefM3Hud2PcSYZqvOW1ldW19I79Z2Nre2d3T9w+aSZRyl1lu5Ee87dgJ872QWcITPmvHnNmB47OWM76U8daE8cSLwhsxjVkvsEehN/RcWxDV1/VKuVY9M7rGrMsDgzXnfb1olk21jGVQyUAR2WpE+gu6GCCCixQBGEIIwj5sJPR0UIGJmLgeZsRxQp6KM8xRIG1KWYwybGLH9B3RrpOxIe2lZ6LULp3i08tJaeCENBHlccLyNEPFU+Us2d+8Z8pT3m1KfyfzCogVuCX2L90i8786WYvAEBeqBo9qihUjq3Mzl1R1Rd7c+FKVIIeYOIkHFOeEXaVc9NlQmkTVLntrq/ibypSs3LtZbop3eUsacOXnOJeB<latexit sha1_base64="ZjurJECuvVkV83/9l19FMtnMVOw=">AAAC1HicjVHLTsJAFD3UF+Kr6tJNIzFhRYrBKDsSNy4xsUAChLRlwIa+Mp2SEMLKuPUL3OovGf9A/8I7Y0lUYnSatmfOvefM3Hud2PcSYZqvOW1ldW19I79Z2Nre2d3T9w+aSZRyl1lu5Ee87dgJ872QWcITPmvHnNmB47OWM76U8daE8cSLwhsxjVkvsEehN/RcWxDV1/VKuVY9M7rGrMsDgzXnfb1olk21jGVQyUAR2WpE+gu6GCCCixQBGEIIwj5sJPR0UIGJmLgeZsRxQp6KM8xRIG1KWYwybGLH9B3RrpOxIe2lZ6LULp3i08tJaeCENBHlccLyNEPFU+Us2d+8Z8pT3m1KfyfzCogVuCX2L90i8786WYvAEBeqBo9qihUjq3Mzl1R1Rd7c+FKVIIeYOIkHFOeEXaVc9NlQmkTVLntrq/ibypSs3LtZbop3eUsacOXnOJeB

2.87 GHz

<latexit sha1_base64="ARLakpOE1j6NDwfRbVYBHQN/lcg=">AAAC1HicjVHLSsNAFD3GV31HXboZrIKrMili213BhV0qWCtYkSQdNZgXk0mhlq7ErV/gVn9J/AP9C++MKeii6IQkZ86958zce700DDLF+fuUNT0zOzdfWlhcWl5ZXbPXN86yJJe+aPtJmMhzz81EGMSirQIVivNUCjfyQtHx7g51vNMXMguS+FQNUnEZuTdxcB34riLqyrarlXqNddmwKyN21LofXdllXuGcO47DNHBqB5xAo1GvOnXm6BCtMop1nNhv6KKHBD5yRBCIoQiHcJHRcwEHHClxlxgSJwkFJi4wwiJpc8oSlOESe0ffG9pdFGxMe+2ZGbVPp4T0SlIy7JImoTxJWJ/GTDw3zpqd5D00nvpuA/p7hVdErMItsX/pxpn/1elaFK5RNzUEVFNqGF2dX7jkpiv65uxHVYocUuI07lFcEvaNctxnZjSZqV331jXxD5OpWb33<latexit sha1_base64="ARLakpOE1j6NDwfRbVYBHQN/lcg=">AAAC1HicjVHLSsNAFD3GV31HXboZrIKrMili213BhV0qWCtYkSQdNZgXk0mhlq7ErV/gVn9J/AP9C++MKeii6IQkZ86958zce700DDLF+fuUNT0zOzdfWlhcWl5ZXbPXN86yJJe+aPtJmMhzz81EGMSirQIVivNUCjfyQtHx7g51vNMXMguS+FQNUnEZuTdxcB34riLqyrarlXqNddmwKyN21LofXdllXuGcO47DNHBqB5xAo1GvOnXm6BCtMop1nNhv6KKHBD5yRBCIoQiHcJHRcwEHHClxlxgSJwkFJi4wwiJpc8oSlOESe0ffG9pdFGxMe+2ZGbVPp4T0SlIy7JImoTxJWJ/GTDw3zpqd5D00nvpuA/p7hVdErMItsX/pxpn/1elaFK5RNzUEVFNqGF2dX7jkpiv65uxHVYocUuI07lFcEvaNctxnZjSZqV331jXxD5OpWb33<latexit sha1_base64="ARLakpOE1j6NDwfRbVYBHQN/lcg=">AAAC1HicjVHLSsNAFD3GV31HXboZrIKrMili213BhV0qWCtYkSQdNZgXk0mhlq7ErV/gVn9J/AP9C++MKeii6IQkZ86958zce700DDLF+fuUNT0zOzdfWlhcWl5ZXbPXN86yJJe+aPtJmMhzz81EGMSirQIVivNUCjfyQtHx7g51vNMXMguS+FQNUnEZuTdxcB34riLqyrarlXqNddmwKyN21LofXdllXuGcO47DNHBqB5xAo1GvOnXm6BCtMop1nNhv6KKHBD5yRBCIoQiHcJHRcwEHHClxlxgSJwkFJi4wwiJpc8oSlOESe0ffG9pdFGxMe+2ZGbVPp4T0SlIy7JImoTxJWJ/GTDw3zpqd5D00nvpuA/p7hVdErMItsX/pxpn/1elaFK5RNzUEVFNqGF2dX7jkpiv65uxHVYocUuI07lFcEvaNctxnZjSZqV331jXxD5OpWb33

(c)

PL [a.u.]

2.97 2.87

2.77

MW frequency [GHz]

|BN V| = 12G

|BN V| = 0G

MW frequency (GHz)

2.77 2.87 2.97

PL signal (a.u.)

2gµB

h |BN V|

magnetic interactions with a bath of paramagnetic impurities both inside the diamond matrix and on its surface [54]. A key requirement is therefore to engineer diamond samples with an extremely low content of impurities, as close as possible to a perfectly spin-free lattice, in order to reach long spin coherence times.

Particular attention must be also paid on the distance between the NV centres and the surface since it is well known to strongly affects both the charge state and the coherence time of NV centres due to the induced surface electronic spin bath [55] even if surface treatment and specific chemical terminations can limits these effects [56-58]. We note that the sensitivity can also be enhanced for the measurement of time-varying signals. Such AC sensing protocols rely on dynamical decoupling sequences of the NV spin sensor, which results in a prolongation of its coherence time to a value commonly referred to as T

, which can be orders of magnitude longer than 𝑇

[50].

While a single NV defect provides an ultimate spatial resolution for imaging applications, the sensitivity can be simply improved by increasing the number N of sensing spins. For an ensemble of NV defects, the shot-noise limited sensitivity 𝜂

then scales as

𝜂

∝ 1

𝐶

√𝑁𝑅𝑇

(2)

A challenge in material science is thus to increase the density of NV defects while maintaining good spin coherence properties. However, the gain in sensitivity is partially compensated by a reduced contrast of spin readout. Indeed, NV defects are oriented with equal probability along the four equivalent <111> crystal directions, leading to a decreased sensitivity because only a quarter of NV spins are efficiently contributing to the detected signal, the others producing solely a background photoluminescence. In addition, luminescence from other impurities, such as the neutral NV

defects, further impairs the signal to background ratio. The spin readout contrast then falls typically to 𝐶

≈ 1 % for large ensembles of NV defects [50].

Mitigating this effect requires (i) to achieve preferential orientation of the NV defects during the diamond growth and (ii) to improve the conversion of NV defects in the negatively-charged state configuration.

Besides providing the highest sensitivity to date [59], ensembles of NV defects can also be used for imaging applications [60, 61]. To this end, a sample of interest is commonly deposited directly on top of a diamond crystal, which contains a thin layer of NV centres near the diamond surface. The spin-dependent PL signal from the NV layer is imaged onto a CCD camera in a wide-field detection scheme, with a spatial resolution limited by diffraction (∼500 nm). In the last years, this method has found numerous groundbreaking applications in very different fields of research [12], including NMR spectroscopy [62, 63], biomagnetism [64], geoscience [65], and condensed matter physics [66-69]. Further performance improvements of this technique require to engineer thin diamond layers with a high NV density featuring long spin coherence time and preferential orientation.

To summarize, current challenges in diamond growth to optimize the performance of NV-based quantum sensing include

(i) Tailoring the diamond matrix so that decoherence is as limited as possible. Although important progresses have been obtained through defect engineering or isotopic purification, coherence times are still far from the theoretical T

limit. This is particularly true when the spins of interest are located near the surface or in a diamond crystal with a high nitrogen content.

(ii) Controlling NV density and charge state. Since many sensing applications rely on dense NV

ensembles to improve sensitivity, controlling the ratio between NVs and other N-containing

defects is crucial. In addition, the close environment of the defect has to favour the occurrence

of the negative charge state with respect to the neutral one.

(iii) Spatially localizing NV centres. The precise positioning of single or ensemble of NV centres both in-depth and in-plane is of importance for incorporating them into cavities or nanostructures, or for improving the performances of wide-field imaging with NV ensembles.

(iv) Controlling NV orientation. Promoting preferential orientation is desirable to limit background noise level, increase sensitivity and simplify device operation.

These different aspects about the material fabrication will be discussed hereafter.

3. The synthesis of “quantum grade” diamond films and crystals

3.1. HPHT grown diamonds

HPHT is well established to produce bulk single crystals with up to a few millimetre thick and millimetre square size that are available commercially in particular for cutting tools applications. This technique typically uses a bath of melted transition metals (such as Co, Fe, Ni, Cr, Mn etc.) in which carbon (in general diamond powder or graphite) is dissolved and re-precipitated on the facet of a small seed in a region of slightly lower temperature (20 to 50 °C less) [70]. This temperature gradient approach involves pressures and temperatures above 5 GPa and 1300 °C respectively. Different heavy set-ups exist that differ from the way the pressure is applied to the cell, such as uniaxial compression with belt and toroid systems, or multi-anvil systems (so- called bars and cubic presses) [71]. This equipment is mostly operated by industrial players (Element Six, General Electrics, Sumitomo, New Diamond Technology etc.) and is essentially destined to mechanical applications for which requirements on purity and quality are moderate. Under adapted and stable conditions though, large crystals can be produced with potentially extremely low extended defect content.

For instance inclusion-free single sector diamonds with stacking faults and dislocation content below a few hundred per cm² have been demonstrated [72, 73]. However the precise recipes developed to reach this degree of perfection are usually a well-kept industrial secret, while the associated costs can be tremendously high (up to several k€ for a 500 µm thin slab). In contrast, standard HPHT crystals typically contain dislocation densities of the order of 10

-10

cm

and visible growth sectoring [74], but their cost is limited to a few hundreds of euros depending on size, orientation and polishing.

Although bulk crystals with low dislocation density can be grown by HPHT, the technique is not well adapted to produce films with a high purity or a controlled doping as required by quantum applications.

Indeed, the high pressures and high temperatures required in this process make the control of possible

contamination coming from impurities trapped in porosity extremely difficult. Standard HPHT diamonds are

usually labelled as type Ib due to the presence of a large amount of non-intentionally doped nitrogen in them

(typically 10-300 ppm) that lead to a yellowish colouration and obvious growth sectoring (see figure 2(a))

[75]. Nitrogen uptake depends on the solvent-catalyst used and its solubility in them. While quantum sensing

requires incorporation of NV centres, it should be emphasized that nitrogen content in HPHT crystals is not

necessarily in the form needed for an optimized sensor. A significant fraction of N is for example present as

substitutional (N

), known as P1 centre. Aggregated forms also exist due to nitrogen mobility being activated

under high pressures and temperatures. The N-V-N or H3 centre is for example frequently created and leads

to emission at 503 nm in PL. Other aggregated forms are less prevalent but commonly found like A-centres

(2 neighbouring N

) or B-centres (N

V) [75]. Their concentration can be of the order of several ppm depending

on the growth conditions or treatment that they have underwent. The addition of getters (Ti, Zr, Al etc.) can

reduce the amount of incorporated nitrogen by preferentially associating and precipitating it as a nitride

allowing fabrication of type IIa colourless diamond crystals by HPHT (see figure 2(b)) [76]. This material leads

to lower background PL and narrower diamond Raman peaks as illustrated in figure 2(c). However in general,

N content cannot be suppressed completely and remains of the order of 0.1 ppm. Growth rates under such

low nitrogen conditions are also strongly reduced which increases the overall cost of the HPHT diamonds [73].

Figure 2. (a) Optical and PL images of a type Ib HPHT substrate. (b) Optical and PL images of a type IIa HPHT substrate. (c) Raman/PL

spectra of type Ib and IIa HPHT substrates obtained with an excitation line at 473 nm. The inset shows the linewidth of the diamond Raman peak. The type IIa diamond was provided by New Diamond Technology (NDT).

Other impurities may also be incorporated in significant amounts including solvents from the melted baths (Ni and Co) or element contaminations (B, Si, Ge) which can lead to the appearance of specific defects or colouring [77]. Around 50 ppm of boron have been measured in some of the purest type IIa diamonds [78]. Boron is known to stabilize the neutral charge state of NV centres and is usually not desirable (see part 4.4). Impurity incorporation dependence on crystal orientation is also an important issue that is associated with variations in colouring and/or luminescence under UV light (figure 2(a)). Incorporation efficiency in (111) growth sectors is usually 2-3 times higher than in (100) and (110) sectors [79]. Isotopic purification of HPHT diamonds to change the

C/

C ratio has been achieved using pyrolytic carbon powder [80] but it remains relatively difficult and uncommon due to the high cost of the precursors and low flexibility of the technique. While residual impurities are difficult to avoid with this synthesis process, intentional additions of certain metals to the bath/catalyst mixture can be explored to create specific colour centres. For example, SiV, GeV or SnV centres which are also interesting systems for QTs have been obtained [81, 82].

This is an important advantage of the HPHT approach because such elements cannot always be easily brought in through the vapour phase or incorporated using the CVD production technique due to limited solubility.

The ability to control crystal morphologies through tuning of the growth temperature and the solvent

has been highlighted and opens the way to obtaining various crystal habits from cubic to octahedral. Control

of the morphology opens the way to the fabrication of larger plates with specific orientations that can be

extracted from such stones [83]. In particular [111]-oriented diamonds can be cleaved from octahedral shape

crystals and are particularly suited as substrates for CVD overgrowth with oriented NV centres (as will be

discussed in part 6) [84]. Bulk crystals can thus be obtained through the HPHT technique with a high

crystalline perfection but limited purity. Nevertheless NV ensembles in type Ib HPHT diamonds have been

studied and exploited for QT demonstrations. Some examples include magnetometry [59], MW photon

storage [85], coupling to superconducting resonators [86], quantum memories [87], hyperpolarisation of

C

[88] or data storage [89]. Although one can benefit from a bulk material that is easily available, the nitrogen

density in the form of substitutional defects is usually a limiting factor and reduces the coherence times (T

)

to typically 1-2 µs only at room temperature [90]. Nevertheless, by reducing the spin bath surrounding NV

centres through isotopic purification and limited nitrogen doping, as well as irradiating the crystal to convert

N

into NV, T

can be extended to several tens of µs [85]. In general electron irradiation followed by annealing has become a rather standard treatment to improve the performance of such HPHT crystals (see part 4.1).

Although some attempts have been made to explore bulk HPHT diamonds crystals in the field of quantum sensing, this material in general fails to provide the purity and the flexibility of fabrication that is required for highly efficient devices. The most common approach thus relies on using HPHT crystals as the starting substrates onto which films with the desired properties are overgrown by CVD.

3.2. CVD grown diamonds

Microwave plasma assisted CVD has become a key technology showing great potential to produce engineered films with the desired doping, isotopic purity and dimensions. The use of CVD-grown diamond films and crystals is relatively wide-spread in QTs although their availability remains scarce. Unlike HPHT, the CVD technique mostly involves academic research groups while commercial availability of high purity plates is limited to a few industrial companies only (Element Six, Diamond Materials, IIa Technologies etc.). A large market for CVD diamond plates is yet to be found. The fabrication of high-quality thick crystals is also technologically challenging with difficult scaling-up which contributes to increase the fabrication costs.

Nowadays, HPHT still remains dominant when it comes to producing bulk synthetic diamonds while CVD is mostly focused on producing thinner layers. Although the size and thickness of the produced crystals is not such a limiting factor for QTs, diamonds may need to be thick enough to be processed and properly oriented or separated from their substrate.

The CVD technique operates at pressures lower than atmospheric pressures (10-300 mbar), under conditions at which graphite should be the thermodynamically stable phase [91-93]. It involves kinetically stabilizing diamond through the production of atomic hydrogen within a high temperature plasma media that preferentially etches away weak sp

bounds, allowing the addition of carbon to the diamond lattice of the substrate. H

and CH

are used in a typical proportion of 95-99 % to 5-1 % respectively. Addition of O

in a small amount (< 2%) is sometimes used in order to increase the etching effect and limit impurities incorporation or non-epitaxial defects formation [94, 95]. In general, activation of the gas is performed through applying a 2.45 GHz MW field to a resonant cavity reactor (figure 3(a)) [96, 97]. Operation under higher pressures (> 100 mbar) and microwave powers (> 2 kW) leads to the formation of a localized plasma region in the core of which temperatures may reach up to 3000 K which is favourable to produce precursors for growth [98]. Indeed, thermal dissociation of the molecules into a variety of atomic and radical species is highly pronounced and may be of up to several tens of percent [99]. Growth is carried out at a temperature in the range 700-1100°C on a diamond seed through either cooling or heating the substrate holder depending on the power density applied. Several providers commercialize MW plasma assisted systems with varying characteristics (Cornes Technologies (Seki systems), Plassys, iplas, optosystems etc.) but a large number of research groups have developed their own equipment. The main differences in those systems are essentially in the way the MW radiation is coupled to the resonant cavity (electromagnetic modes), the location of dielectric windows as well as the design of the holder (translatable, rotatable, cooled or heated). High-power operating reactors are preferred for achieving high growth rates and low defect bulk diamond crystals.

However, the low-power regime may be advantageous for ensuring nm-scale control over the thickness of the layers and a precise positioning of dopants or colour centres (see part 5.1).

With a hetero-substrate (like a silicon wafer) a polycrystalline film in which grain size directly depends

on thickness through a columnar growth mode is generally obtained. Under certain growth conditions, films

may exhibit a particular texture or preferential orientation [100]. The presence of grain boundaries is

however deleterious to obtaining long coherence times and low background luminescence. Polycrystalline

diamonds are usually not preferred for sensing applications. Interesting properties for single NVs have

however locally been found within the larger grains of polycrystalline films [44, 101]. They also offer the

advantage of providing a large and flexible platform for processing them into photonic crystals and resonators that would be highly desirable for QTs [102].

Figure 3. (a) Microwave Plasma Assisted CVD system allowing the growth of diamond. The inset shows a zoom into the plasma region

in which a diamond is positioned for growth. (b) High purity thick CVD diamond layer grown on a yellow HPHT diamond substrate and (c) freestanding CVD diamond film obtained after removing the HPHT substrate by laser cutting and polishing. (d) PL image of a high purity thick freestanding CVD diamond film showing only very weak blue luminescence in the corners corresponding to the presence of stress. (e) PL image of a N

doped thick freestanding CVD diamond film showing orange luminescence corresponding to the presence of both NV

and NV

colour centres.

Heteroepitaxial growth on specially developed templates that include a thin monocrystalline iridium layer deposited on an oxide thin film on silicon or a bulk crystal (Yttria stabilized Zr0

, a-plane Al

O

or SrTiO

) is another possible approach that promises wafer-scale deposition area [103-105]. In recent years, some important efforts in material development have been devoted to obtaining higher quality and larger films with new venture companies starting to commercialize them (Audiatech, Namiki etc.). The complex steps that lead to oriented diamond growth include deposition of the epitaxial Ir films on an appropriate substrate, the biased enhanced nucleation of diamond domains on them, thickening of the film and limitation of dislocation density through patterning of the surface [106]. In general dislocation densities still remain high even in thick films (> 10

cm

²) [107] and impurities such as silicon are hard to avoid. Nevertheless a recent assessment of state-of-the-art films produced through this approach have demonstrated T2 coherence times of 5 µs supporting the idea that they may provide a useful larger platform for future applications in QTs providing material quality and availability are improved [108].

Homoepitaxial growth onto a diamond seed (generally a type Ib HPHT substrate) is the preferred fabrication route for obtaining high quality and purity diamond films that are suitable for QTs (figure 3(b) and 3(c)). Although CVD is relatively simple in its operating principle, obtaining layers with a given defect concentration and the desired thickness has animated a large number of research activities through the past decades, in particular within the earlier and demanding context of power electronics. Substrate selection and preparation plays an important role in the epitaxial overgrowth and adapted polishing or etching of the surface prior to growth helps limiting the propagation of defects from the interface [109]. Maintaining constant growth conditions, especially temperature, during long periods of time is also a limiting factor when thicker layers are desired. The development of specific substrate holders that include cooling with gas mixtures or vertical translation may be needed [110-112]. The presence of uncontrolled amounts of N

or O

from reactor leaks or impure feed gases have important consequences and can induce the formation of

polycrystalline defects that would quickly ruin the entire growth run [91, 98, 113, 114]. To this end, care must be taken to frequently check for potential leakage sources and to use dedicated purifying systems especially for hydrogen. When a good control of gas environment is successfully achieved, single crystal diamond plates with good purity or intentionally doped with a controlled nitrogen amount can be prepared (see figure 3(d) and 3(e)). Other potential contamination sources may be released by the constitutive materials of the reactors themselves (metal walls, quartz windows, molybdenum holders etc.) such as boron, silicon or nitrogen. Besides choosing adapted materials for the reactor furniture, the design should ensure that internal parts are appropriately cooled down or positioned far away from the high temperature plasma media. The appearance of SiV emission in CVD diamonds is nevertheless very common and is even used as a criteria for establishing diamond’s synthetic origin in gemmology [115]. Finally it should be noted that hydrogen, one of the main elements involved in the growth process is usually overlooked although it is one of the main impurities in CVD-grown crystals. Hydrogen-vacancy defects known as H1 centres are paramagnetic and show-up in EPR together with the nitrogen-vacancy-hydrogen (NVH) for example [116, 117].

The ability to prepare isotopically enriched layers is also a particularly useful asset of CVD grown diamond films. Indeed growth from methane using natural isotopic carbon ratio leads to the presence of 1.1 % of

C in the films which is a non-zero nuclear spin element. Coupling of the NV spins to nearby

C atoms is the main source of decoherence for films with a low NV amount (< 0.1 ppm) far away from the surface [54, 118]. Reducing the amount of

C is relatively straightforward by substituting the conventional methane source with an enriched

C methane cylinder. By doing so, T

times have been successfully extended from a typical value of 0.5 ms up to a record of 2.4 ms [119]. Nevertheless the cost of such sources is several orders of magnitude higher than a standard methane cylinder. Moreover the specifications in terms of N

or CO

background content are usually much higher than high-purity grade methane and may require additional purification steps with dedicated purifier cartridges. On the other hand, intentional addition of

C in CVD-grown films can be achieved to deviate from the natural isotopic ratio. Particular schemes have been proposed that explore coupling of a NV spin to a nearby long-lived nuclear spin to further extend quantum storage times [120]. Dynamic nuclear polarization may also be useful to magnetic resonance spectroscopy and imaging applications [121].

In general, one of the main advantages of the CVD growth approach for making “quantum grade”

diamonds is the ability to engineer stacked layers of different doping and composition in a dynamic and very flexible way. Indeed the gas phase environment can be controlled to an extremely high level while changing from one composition to another can be done with abrupt interfaces providing residence time of gas species are taken into account (see part 5.1). CVD diamond fabrication of specially designed bulk crystals or thin films has thus become a cornerstone of the developments that the quantum technologies based on this material system have witnessed.

4. Creating colour centres with good coherence properties

4.1. Implanting colour centres in high purity CVD diamonds

While the CVD technique allows the fabrication of isotopically enriched diamond films with extreme purity, a varying amount of colour centres need to be incorporated within this matrix to provide the sensing functionality. A widely followed approach consists of locally implanting nitrogen ions (N

) or a molecule containing nitrogen (N

, CN

etc.) in “electronic grade” (i.e. high-purity and non-luminescent) CVD diamonds.

In general 3 steps are required: (i) introducing impurities, (ii) creating vacancies (that may be co-implanted

together with the impurity or afterwards), (iii) annealing to heal defects and diffuse vacancies so that the

complex defect can be formed. The present paper does not intend to give a detailed review of the

optimization of implanted colour centres in diamond and readers are advised to refer to the following articles [122, 123], however some of the main trends are presented below.

Regarding the first step, the ions to be implanted can be accelerated in a wide range of energies from typically 2 keV to 20 MeV leading to implantation depths of 3 nm to about 5 µm respectively. This obviously requires rather different implantation set-ups from small table-top sources for low energies to large tandem accelerators to reach the MeV regime. It allows creating specific luminescent patterns within the diamond substrate as illustrated in figure 4(a). It should be noted that the ion energy not only influences the penetration depth of the ions but also the creation yield of NV centres with respect to each N atom entering the diamond lattice [124]. In fact, the higher the energy, the higher the number of vacancies that are co- created leading to higher yields. Typically, values range from 0.1 % at 2 keV up to about 45 % at 18 MeV.

However at high acceleration energies, spatially positioning the implanted ions with accuracy becomes difficult due to the statistic distribution of collisions with atoms in the lattice that leads to a lateral and depth spread called straggling. Getting a spatial accuracy of less than 5 nm for example requires that ions are accelerated to an energy below 10 keV which limits penetration to 10 nm only (i.e. to near-surface NV centres). This shows that a trade-off exists between high-yield high-depth NVs and low-yield low-depth but highly localized NVs, depending on the energy of the incoming ions [125]. To go beyond those limits, strategies have been developed to increase the positioning accuracy by implantation through a pierced AFM tip [126], mica channels or opened PMMA masks [127].

An additional advantage of the implantation technique relies on its ability to generate defects from elements that cannot be easily grown-in directly by CVD due to too high steric hindrance, low stability or difficulty in bringing them through the gas phase. Besides, co-implantation with other elements brings additional flexibility in the generation of complex defects. Lühmann et al. have for example studied a wide variety of colour centres that can be created through implantation of elements as varied as Mg, Ca, F, O etc.

in a matrix that already contains other implanted impurities of phosphorous or boron [128]. This obviously leads to an exhaustive variety of combinations and adds additional complexity to this approach in determining the most relevant colour centre for a given application. Control of the charge state of created vacancies has been accomplished by implanting nitrogen into n-type doped diamond rather than in a standard intrinsic crystal. In this way, the negative charge state of the vacancies is promoted which reduces their clustering and thus increases the probability that they associate to a single nitrogen atom to form a NV centre. Record creation yields of about 75 % have been reported in sulphur doped diamonds [129].

Introducing additional vacancies into the diamond crystal can also be explored in order to boost the

NV/Ns ratio (creation yield). Such irradiations are accompanied by the creation of the GR1 luminescent defect

(neutral vacancies) which intensity depends on the dose, the energy and the type of ions that are used

(figure 4(b)). Helium ions accelerated to a few keV provide for example a way to locally create vacancies at a

controlled depth of a few tens of nm and thus the ability to generate so called delta-profiles [130]. He

ion

beams can also be focused down to a small size to create patterns [131]. However, evidence exist of the

creation of specific colour centres related to the implantation of helium atoms within the lattice. The optical

properties of such Helium-Vacancy (He-V) centres have recently been studied [132]. Other irradiations using

protons or electrons present the advantage of having a lower mass as compared to helium. Their stopping

range is much longer which allows for a more uniform creation of vacancies through the volume of the

sample. Electron irradiation at several MeV and doses of the order to 10

-10

cm

has become a standard

treatment to Ib HPHT diamonds in order to increase NV density [133]. Local irradiation at lower energies

(around 200 keV) has also proved successful using the electron beam of a Transmission Electron Microscope

(TEM) [134]. At this energy the penetration depth can be estimated to about 140 µm. Vacancy creation

efficiency is also more limited with a minimum energy for vacancies creation of about 145 keV. Nevertheless

this technique provides a way to generate local NV patterns [135, 136]. An alternative approach is the creation of vacancies through ultrafast (fs) laser irradiation pulses. Single NV centres can thus be written locally into arrays with a positioning accuracy of about 200 nm and coherence times of several hundreds of µs equivalent to naturally occurring NVs [137, 138].

Figure 4. PL image of a high purity freestanding CVD diamond film after localized nitrogen ion implantation which leads to the

appearance of orange/red spots corresponding to NV

and NV

colour centres (Collaboration University of Leipzig). (b) Evolution of vacancies/ion number and GR1 fluorescence as a function of implanted ion mass. (c) Evolution of the fluorescence of defects created by implantation as a function of annealing temperature. (b) and (c) are adapted from Lühmann et al [128].

Annealing ion implanted diamonds is a key ingredient to increase NV density and improve their coherence properties. This step allows vacancies to diffuse and defects to heal so that highly coherent NV centres are formed successfully following irradiation. Since vacancies in diamond are mobile above 700 °C, typical annealing temperatures after irradiation are in the range 800-1000 °C with the treatment carried out for a few hours (1-10 h). This is illustrated in figure 4(c) where the number of NVs is seen to increase with temperature together with a decrease of GR1. Annealing simultaneously when doing the ion implantation provides a way to reduce collateral damage and preferentially associate the vacancies with a nearby N rather than forming clusters [139]. Annealing at too high a temperature (> 1200°C) is likely to lead to the formation of vacancy clusters or to the thermal dissociation of NVs which should be avoided. Nitrogen atoms can also become mobile at temperatures of the order of 1600 °C possibly forming complex clusters like H3 (N-V-N) as shown in figure 4(c). In general T

times of implanted and annealed NVs remain below those of originally presents in the diamond by 1-2 orders of magnitude (typically 1-10 µs) due to the presence of other defects and residual damage that cannot be completely annealed out. Optimized annealing treatments at higher temperatures [140, 141] or composed of successive steps with various annealing temperatures and durations have been proposed to obtain NV coherence times close to those of native NVs [142]. Nevertheless there probably does not exist a universal efficient annealing step, the optimized procedure strongly depending on the initial quality of the diamond as well as the starting NV density.

4.2. In situ doping of colour centres

While ex-situ creation of colour centres is a flexible approach with accurate positioning ability, it

generally does not allow obtaining defects with as good coherent properties as naturally occurring ones from

in-situ doping. Intentional doping during CVD growth can indeed be achieved by injecting into the plasma a

precursor gas containing the element to be doped. N

is the most widely used dopant for NV doping. This

molecule has a very strong bond energy (9.8 eV) which requires high plasma power densities to efficiently

dissociate it. Nitrogen doping into the tight diamond lattice is also not energetically favourable. Both those

issues lead to low doping efficiencies of about 10

to 10

[143, 144]. When low plasma power densities are

used (< 1 kW MW power), several % of N

are usually required to reach N

concentrations of the order of 0.1 ppm [145, 146].

An additional limitation is that only a small fraction of the total nitrogen will be incorporated as a complex associated to a vacancy, with the main part being substitutional to a single carbon atom (N

). A typical yield (NV/N

) for untreated as-grown CVD diamonds is of the order 1/300 or below [147]. Therefore it should be highlighted that the creation yield for in situ doping is similar to that obtained for nitrogen implanted at medium energies. This is a particularly limiting factor since not only the amount of NV centres to be used for sensing will be limited but the large proportion of nitrogen, a paramagnetic impurity, will induce decoherence especially for the highest doping levels [118]. In addition, a small part of the total nitrogen in the CVD-grown diamond, more or less the same proportion as that of NVs may occur in the form of NVH complex [148]. These defects are fairly common in CVD diamonds grown under high nitrogen additions since hydrogen is one of the main ingredients for CVD growth. They can be detected in their negative form as a line in EPR and are also visible as a sharp absorption at 3123 cm

in FTIR in their neutral charge state [149, 150] (figure 6(c)). While hydrogen impurities are likely to passivate part of the NV centres and produce additional magnetic noise, these complexes cannot be easily annealed out even at very high temperatures. It would be desirable to improve the NV yield by changing the growth conditions such as substrate orientation, pressure and MW power, gas phase composition (methane, hydrogen, nitrogen and oxygen). However, no systematic study exists so far on the influence of growth conditions on the NV creation yield most likely due to the difficulty in accurately measuring the concentration of those defects in thin diamond films.

While N

is the most frequently used dopant, other molecules have been shown to lead to improved NV doping efficiency and photostability. For instance, N

O which has a much lower dissociation energy is also available as a high-purity gas. While high-density NV ensembles (around 10 ppb) created through addition of N

are subject to blinking and charge state instability especially under high laser pumping power (figure 5(a- c)), those formed from N

O are much more stable (figure 5(d-f)) [151]. This is possibly related to the presence of a low amount of oxygen near the growing surface due to N

O dissociation in the plasma that etches away any defects that act as traps for charge carriers [152]. However, the use of other dopant sources is not very widespread and would require more dedicated studies.

Figure 5. (a) and (b) PL images of a CVD diamond film grown with 100 ppm of N2

in the gas phase obtained with a confocal scanning

microscope (laser, 532 nm), allowing estimating the NV

density to 10 ppb and showing the poor photostability under high laser

power. (c) ODMR spectra obtained under a magnetic field of 3 mT and where the characteristic hyperfine splitting is observed with

a FWHM of 0.58 MHz. (d) and (e) PL images on a CVD diamond sample grown with 500 ppm of N

O in the gas phase allowing

estimating the NV

density to 15 ppb and showing improved photostability even under high laser power. (f) ODMR spectra obtained

Besides NV, other colour centres can be introduced in-situ in CVD-grown diamonds. However, as compared to the HPHT process or to ex-situ implantation, the introduction of a wide variety of impurities is relatively limited. SiV and GeV centres have been obtained through addition of a varying amount of silane or germane gases (SiH

and GeH

) [153, 154]. For those elements though, doping using a solid-state source (like a small piece of Si or SiC placed near the growing diamond) is usually the preferred approach due to its simplicity and non-toxicity [155, 156]. Dopants that modify the Fermi level of the semi-conducting diamond crystal such as phosphorous (n-type) or boron (p-type) can also be advantageously used to tune the charge state of the colour centres (see part 4.4). This is generally achieved through additions of TBP (tri-butyl phosphine), TMB (tri-methyl boron) or B

H

(diborane) [157]. Although it is not directly involved in the creation of specific colour centres, the addition of O

during growth is also sometimes explored to improve the crystalline quality of the films thereby potentially improving the coherent properties of in situ doped NV centres [158].

4.3. Controlling colour centres density

By in-situ doping, NVs can be created in a wide range of doping levels from isolated single centres to ensembles of several 10’s of ppb without any post-treatment simply by tuning the added gas concentration during growth. For some nanoscale sensing applications, or for quantum memories [159] the manipulation of single NVs with long coherence times is required. Extremely low amounts of NVs (or even more, no NV at all) are however particularly difficult to achieve and require carefully purified gas sources and leak-tight reactor chambers. While high-purity “electronic grade” commercial diamonds (specified as N

< 5 ppb) do not normally display luminescence originating from NVs, it has been shown that after annealing at high temperature (1600 °C for 4 h), vacancies diffuse and are able to associate to nitrogen [128]. This leads to up to 1 NV/µm² and indicates that even carefully prepared CVD diamond films may still contain a low but non- negligible residual background of nitrogen. The creation of isolated NVs (0.1-1 ppb) relies on low additions of N

during growth in general diluted in hydrogen (around 0.1-10 ppm) and thus requires a precise control of the gas phase composition to achieve the desired concentration [143]. Doping efficiency also strongly depends on growth parameters such as power density, temperature and substrate orientation and can thus vary on different set-ups.

On the other hand, high density NV ensembles are desirable for many quantum sensing schemes [160]

since the sensitivity of a given sensor will depend on the square root of the number of sensing spins [13] (see

part 2, eq. 2). However, obtaining very high NV concentrations (> 100 ppb) is problematical by direct CVD

growth since large additions of N

, above typically 250 ppm under high power density conditions, is

accompanied by a degradation of the surface morphology [161] particularly at the edges of the crystal [162],

even leading to a total loss of epitaxy for the highest levels leading to the appearance of twins and

polycrystalline material. Nitrogen solubility is also limited in CVD and cannot allow reaching as high levels as

those typically reached in type Ib HPHT diamonds (a few tens or hundreds of ppm). Under moderate nitrogen

doping levels though, CVD diamonds often exhibit a brown colour indicative of the presence of vacancy

clusters or dislocations as shown in figure 6(a) [163]. This is due to large absorption features at 510, 360 and

270 nm, the latter being directly correlated with N

(figure 6(d)). Very high brightness diamond films can be

obtained by such doping (figure 6(b)). For thick crystals, N

concentration can be directly evaluated from the

intensity of the 1344 cm

feature in FTIR as shown in figure 6(c) [164]. It has been proposed that addition of

a low amount of oxygen during growth together with nitrogen helps limiting the formation of such defects

and the appearance of the brown colour [95]. Highly N-doped CVD diamonds nevertheless contain a large

amount of residual defects that strongly reduce their optical properties. HPHT or LPHT post-treatments

(> 1500 °C) are often required to improve their colour due to a partial rearrangement or annihilation of point defects [165].

Figure 6. (a) Optical images of thick freestanding CVD diamond samples grown with 20 ppm, 40 ppm and 100 ppm of N2

in the gas phase from left to right, ranging from colourless to dark brown. (b) Corresponding Raman and PL spectra performed on these samples showing high emission from NV

and NV

luminescence at 575 nm and 637 nm respectively. (c) FTIR spectrum carried out on a thick freestanding CVD diamond sample grown with 500 ppm of N

O in the gas phase allowing to clearly identify and quantify N

, N

and NVH

defects. (d) UV-Visible absorption spectrum of a high purity CVD diamond plate and a CVD diamond plate grown with 500 ppm of N

O in the gas phase. Absorption at 270 nm is due to N

while other broad absorption bands are related to vacancy clusters.

Besides the difficulty in growing CVD crystals in the presence of a high concentration of nitrogen in the gas phase, an additional issue comes from the fact that NVs represent only a small fraction of the total incorporated N content (see part 4.2). The contribution of

N nuclear spin bath on NV spin’s dephasing starts to overcome that of natural isotopic

C for concentrations above 0.1 ppm. At 10 ppm total nitrogen, T

times drop to about 10 µs [50]. Therefore this leads to a trade-off between high NV density and long coherence times. In order to circumvent this, partial conversion of N

into NVs can be obtained through an appropriate irradiation using high energy electrons [59] or He

ions (see part 4.1). Using the later, Kleinsasser et al. [166]

achieved NV

densities of the order of 1 ppm which is only 10 fold lower than the highest densities reported in irradiated Ib HPHT diamonds [167] while ESR linewidths remained narrow (200 kHz). N to NV conversion rates of the order of 10-20 % are possible through an appropriate irradiation [168]. In this case dipolar interactions between proximal NV centres might dominate the dephasing rather than NV to N coupling.

4.4. Controlling NVs charge state

The NV centre possesses neutral and negative charge states with zero phonon line emissions at 575 and 637 nm respectively (see part 2). Both are usually present in nitrogen-doped single crystal diamonds.

Quantum sensors exploit the optical properties of the negatively charged NV centres (spin S = 1) and

therefore the neutrally charged centres with spin S = ½ are undesirable. They may lead to overlapping of the

PL emission due to broad phonon side bands as well as magnetic noise that degrades spin coherence times.

It is generally admitted that NV centres acquire their negative charge from nearby electron donors. One obvious candidate for this charge transfer is the substitutional nitrogen that represents a large fraction of the total nitrogen content in non-irradiated nitrogen-doped HPHT and CVD diamonds [149]. For this reason, under a low power excitation in the range 450-610 nm to limit photo-ionization, the steady-state NV

centre population is typically about 75 % of the total NV amount [169]. However, this value can vary depending on the nitrogen doping level in the diamond crystal. Type Ib HPHT diamonds with high nitrogen content (several tens of ppm) may have a higher proportion of NV

[170, 171]. For example, in the PL spectra of figure 7(a) with an excitation at 532 nm, NV

emission is almost undetectable as compared to that from NV

due to the large amount of nitrogen (circa 100 ppm) in this electron irradiated type Ib crystal. To the contrary, when too high a proportion of N

are converted to NVs by irradiation, there may not be a sufficient amount of donors close enough to NVs to provide the necessary electron. In this case, emission from NV

tends to saturate while that from NV

increases, which occurs above a certain irradiation dose [172].

Promoting NV’s negative charge state can also be achieved with shallower electron donors than nitrogen such as phosphorous or sulphur [129, 173]. Boron which is an acceptor impurity produces an opposite effect by favouring the neutral charge state. Groot-Berning et al. have clearly shown the effect of co-doping in implanted diamond films [174] as illustrated in figure 7(b). Fermi level tuning can also be achieved with in- situ doped CVD diamond films [175] allowing a fine control over NV’s charge state. Intentional doping of diamond by phosphorous during CVD growth is however particularly challenging due to the low doping efficiency of this element into the diamond while n-type conductivity is limited by compensating defects and high activation energy (0.6 eV) [176]. Nevertheless, electrical control over NV’s charge state (and emission) has been shown with p-i-n junctions and switching from NV

to NV

has been obtained by applying a strong bias [177].

Regarding shallow NVs, their charge state also strongly depends on the chemical species present at the diamond’s surface (typically hydrogen or oxygen termination, see figure 7(b) and 7(c)). Since hydrogen termination of diamond films induces a 2D hole gas by a surface transfer doping mechanism [178], this termination is not favourable to negatively charged NVs. On the contrary, oxidative etching of the diamond by heating at temperatures around 450 °C in an O

atmosphere leaves the surface oxygen-terminated and promotes NV

[179]. Active tuning of the band bending and thus of NV’s charge state, through an electrolytic gate electrode [36] or using a Schottky metal contact [180] have been demonstrated leading to on-demand switching of the PL emission. In general, it should be noted that O-terminated diamond surfaces, which are easily obtained through either acidic treatment (typically by dipping in boiling H

SO

/HNO

mixtures), by exposure to a soft microwave oxygen plasma or UV-ozone lamp, are preferred since they provide a more chemically stable environment for NV

centres even if some more surface terminations based on Fluor or nitrogen should be more mentioning since they lead also to better coherence times [57, 58, 181].

Besides those effects intrinsic to the diamond environment and doping, the magnitude of NV

and NV

emission in PL also depends on the excitation wavelength of the laser due to different absorption cross-

sections [171]. Figure 7(a) shows that the NV

/NV

ratio is strongly affected by changing the laser from 532 nm

to 473 nm for a diamond irradiated with an electron dose of 1.5x10

cm

. It has indeed been shown that

excitation in the blue is more favourable to NV

. There is near equal excitation at 514 nm and in this case,

the strengths of the zero-phonon lines is a good indicator of the relative concentrations of the two NV charge

states [171]. In order to more precisely measure the charge state ratio, decomposing the PL spectrum

obtained with a 532 nm excitation has also been proposed as a more straightforward approach [182].